Environmental Research Letters

Environ. Res. Lett. 9 (2014) 024005 (9pp) doi:10.1088/1748-9326/9/2/024005

Rapid and extensive warming following

cessation of solar radiation management

Kelly E McCusker

1, Kyle C Armour

2, Cecilia M Bitz

3and David S Battisti

3

1 School of Earth and Ocean Sciences, University of Victoria, Victoria, BC, Canada2 Department of Earth, Atmospheric and Planetary Sciences, Massachusetts Institute of Technology,Cambridge, MA, USA3 Department of Atmospheric Sciences, University of Washington, Seattle, WA, USA

E-mail: kemccusk@uvic.ca

Received 14 July 2013, revised 22 January 2014Accepted for publication 24 January 2014Published 17 February 2014

Abstract

Solar radiation management (SRM) has been proposed as a means to alleviate the climateimpacts of ongoing anthropogenic greenhouse gas (GHG) emissions. However, its efficacydepends on its indefinite maintenance, without interruption from a variety of possible sources,such as technological failure or global cooperation breakdown. Here, we consider the scenarioin which SRM—via stratospheric aerosol injection—is terminated abruptly following animplementation period during which anthropogenic GHG emissions have continued. We showthat upon cessation of SRM, an abrupt, spatially broad, and sustained warming over landoccurs that is well outside 20th century climate variability bounds. Global mean precipitationalso increases rapidly following cessation, however spatial patterns are less coherent thantemperature, with almost half of land areas experiencing drying trends. We further show thatthe rate of warming—of critical importance for ecological and human systems—is principallycontrolled by background GHG levels. Thus, a risk of abrupt and dangerous warming isinherent to the large-scale implementation of SRM, and can be diminished only throughconcurrent strong reductions in anthropogenic GHG emissions.

Keywords: climate engineering, geoengineering, solar radiation management, abrupt climatechange

S Online supplementary data available from stacks.iop.org/ERL/9/024005/mmedia

1. Introduction

Stratospheric aerosol injection has emerged as a popular,hypothetical solar radiation management (SRM) technique dueto its technological and economic feasibility and potential toswiftly and effectively cool the planet and avoid impendingclimate emergencies (Keith et al 2010, Vaughan and Lenton2011). Moreover, the observed cooling following volcaniceruptions (Robock 2000) and numerical simulations of SRMwithin climate models (e.g. Rasch et al 2008a, Robock et al

2008, Ammann et al 2010, McCusker et al 2012) serve as

Content from this work may be used under the terms ofthe Creative Commons Attribution 3.0 licence. Any further

distribution of this work must maintain attribution to the author(s) and thetitle of the work, journal citation and DOI.

strong evidence that an enhanced stratospheric aerosol layercould effectively curb global warming. In order to stabilizeglobal climate near present-day conditions, SRM would needto provide a negative shortwave radiative forcing that iscomparable to the current global energy imbalance, observedto be on the order of 0.5–1 W m2 (Hansen et al 2011, Lymanet al 2010). Of course, as greenhouse gas (GHG) emissionscontinue, the magnitude of SRM forcing required to cool orstabilize global climate will correspondingly increase.

If such an enhanced stratospheric aerosol layer wereproduced, any interruption to its continual maintenance wouldcause a quick return to natural aerosol levels within 1–2 years(Rasch et al 2008b). In turn, global temperature would increaserapidly as the climate adjusts to the full, unmasked GHGradiative forcing. Previous evaluations of SRM termination

1748-9326/14/024005+09$33.00 1 c� 2014 IOP Publishing Ltd Printed in the UK

Environ. Res. Lett. 9 (2014) 024005 K E McCusker et al

(Wigley 2006, Matthews et al 2007, Ross and Matthews2009, Brovkin et al 2009, Robock et al 2008, McCuskeret al 2012, Irvine et al 2012) have focused on the globaland annual mean climate response under ‘business-as-usual’GHG emissions scenarios. These studies suggest that therates of global warming following SRM cessation could reach1 �C/decade or greater, far exceeding warming rates had noSRM been implemented. Such a rapid temperature changewould substantially affect human and ecological systems,whose resilience would be limited by rates as small as a fewtenths of a degree per decade (van Vliet and Leemans 2006,Lenton 2011).

Of critical importance for ecosystem adaptation and sur-vival is the geographic structure of warming and the rate ofseasonal temperature change (Lenton 2011). Crop yields, forexample, are highly sensitive to growing season (typicallysummertime) temperature, and have already declined in re-sponse to 20th century warming (Lobell et al 2011). Indeed,major threats to food security have historically been overcomein part by regional food surpluses compensating for low yieldsin other regions (Battisti and Naylor 2009); more spatiallybroad and rapid warming could preclude the existence of suchcompensating regions. Moreover, a high rate of environmentalchange reduces the mean fitness of populations (Bell andCollins 2008), yielding at minimum, a less diverse group ofthe ‘luckiest’ species (Bell and Collins 2008) resulting in lossof biodiversity. Survival of migratory animals depends on thedistance to their optimal climate and the speed at which theycan disperse; many mammals are already at risk of losing pacewith climate change (Schloss et al 2012). Thus, widespreadand rapid warming following SRM cessation could issue aone-two punch to human and ecosystem adaptation.

Motivated by the above assessments of impacts of rapidchange on ecosystems and human systems, we first considerhere the geographic structure of temperature and precipitationchange following SRM cessation, with particular focus on therates of seasonal warming over land, features lost in the globaland temporal averaging of earlier studies. We then analyze theresponse to SRM cessation over a wide variety of plausiblescenarios, spanning a range of climate sensitivities, futureGHG emissions trajectories, and SRM termination years. Theresults are considered in the context of the temperature andprecipitation trends experienced over the past century, to whichecosystems and human systems have become well adapted intheir respective regions.

2. Geographic pattern of land surface temperature

trends

To evaluate the spatial climate response to a SRM termination,we use the Community Climate System Model version 4(CCSM4) (Gent et al 2011), a state-of-the-art fully-coupledgeneral circulation model (GCM). We obtained six CCSM420th century simulation (Historical; 1900–2005) ensemblemembers, 300 years of a Preindustrial control simulation,and a CCSM4 RCP8.5 simulation from the National Centerfor Atmospheric Research. The RCP8.5 simulation is forcedwith GHG and aerosol emissions into the future, such that the

radiative forcing reaches about 8.5 W m�2 above preindustriallevels by 2100 (Moss et al 2010). The Historical simulationshave slightly varied initial conditions and are identically forcedwith historical GHG and aerosol emissions plus volcaniceruptions.

To simulate a SRM scenario within CCSM4, we imposein the year 2035 a latitudinally distributed, zonally uniform,monthly climatology of stratospheric sulfate aerosol concen-tration (as in McCusker et al (2012)). At this time, globalmean surface air temperature (SAT) is about 1 �C higherthan the end of the 20th century (1970–1999 mean) and about2 �C higher than preindustrial SAT. We increase the prescribedsulfate burden from zero to 8 teragrams of sulfur equivalent(TgS) in 3 years to approximately return to the end of the 20thcentury temperature, then increase the concentration to providea roughly equal and opposite radiative forcing to RCP8.5(at a rate of 0.67 TgS/year) thus approximately stabilizingglobal climate. To simulate an abrupt SRM cessation, thesulfate burden is zeroed after 25 years of implementation (year2060). We conduct an ensemble of two such ‘SRM shutoff’simulations with slightly varied initial conditions.

Following SRM termination, the global average annualmean SAT rapidly approaches the temperature had no SRMbeen implemented (figure 1). Global mean SAT increases bynearly 4 �C within 30 years, compared to a SAT increase of lessthan 2 �C in that same time period under the RCP8.5 scenario.The linear trend of global SAT from the 2-member ensemble is1.16 �C/decade over the first 20 years of the Shutoff scenario,consistent with previous findings (e.g. McCusker et al 2012,Irvine et al 2012). This trend is about six times larger thanthe observed global warming trend since 1975 (0.2 �C/decade)(Hansen et al 2006) and sixteen times greater than the observedtrend over the entire 20th century (0.07 �C/decade) (Hansenet al 2006).

We focus here on the spatial pattern of summertimetemperature change in each hemisphere (JJA to the north ofthe equator, and DJF to the south) because of the implicationsfor agricultural productivity, but warming rates for the wintermean and annual mean are consistent with those in summer(see figure S1 available at stacks.iop.org/ERL/9/024005/mmedia). The ensemble land-averaged 5-year summer SATtrend following SRM termination is 3.3 �C/decade, whilelocal trends are as high as 15 �C/decade (figure 2(a)); overmuch of the northern high latitudes, temperature trends arenear 8 �C/decade. The 20-year trends are more spatiallyhomogeneous, with an average trend of 1.25 �C/decade, andmany regions near 2–2.5 �C/decade (figure 2(d)).

To put the SRM Shutoff trends into context, we normalizethe 5- and 20-year trends by the typical variation (standarddeviation) of local simulated historical trends of the sametime interval (figures 2(b) and (e)). The historical standarddeviation is calculated from distributions of 5- and 20-yeartrends sampled from the Historical CCSM4 ensemble (furtherdescribed in supplementary materials available at stacks.iop.org/ERL/9/024005/mmedia). While the 5-year warming trend ison average 1.3 standard deviations of 20th century variabilityover land, there is substantial spatial variation (figure 2(b));in many food insecure regions, such as Sub-Saharan Africa

2

Environ. Res. Lett. 9 (2014) 024005 K E McCusker et al

3

Figure 1. Annual mean, global mean (a) surface air temperature (�C) and (b) precipitation (mm/day) for Historical (black), RCP8.5 (red),average of 4 SRM Ramp simulations (blue line) and ensemble range (light blue shading), two SRM Shutoff simulations (orange), and aPreindustrial control (gray) simulation for reference.

Figure 2. (a) The ensemble average summer 5-year SAT trends (�C/decade) for the period following SRM termination in the Shutoffscenario. (b) The 5-year trends shown in (a), but normalized by the standard deviation of Historical 5-year trends at each grid point. (c) The5-year trends shown in (a), but normalized by the standard deviation of Preindustrial 5-year trends at each grid point. ((d)–(f)) are as((a)–(c)), but for 20-year trends. The white stripe at the equator indicates the discontinuity in season from northern hemisphere to southernhemisphere (JJA and DJF, respectively). Trends over the ocean are shown in figure S2 (available at stacks.iop.org/ERL/9/024005/mmedia).

and South Asia (FAO et al 2012), trends exceed 2 standarddeviations. However, the land-average 20-year warming trendis 4.5 standard deviations of 20th century variability, andlocal trends exceed two standard deviations in the majority ofregions (figure 2(e)). Twenty-year trends are drastically outsideof 20th century bounds within the tropics, where variability isrelatively small (figure S3 available at stacks.iop.org/ERL/9/024005/mmedia), and exceed 6 standard deviations within manyfood insecure regions. Annual mean, land-averaged trendsare 1.8 and 5.6 standard deviations for 5- and 20-year trendsrespectively (table 1), and global averages are further out ofbounds due to relatively low variability over the world’s oceans

(see figure S2 available at stacks.iop.org/ERL/9/024005/mmedia for global summer trend maps).

The Historical simulations were chosen as the comparisonperiod in order to put trends after SRM cessation into thecontext of ‘what humans and ecosystems have experienced’over the previous century, and therefore no detrending wasapplied to the Historical simulations. Trends normalized byPreindustrial control variability, where no time-varying forcingis included, are generally similar but slightly more anomalousthan when normalized by Historical variability, especiallyover 20 years (table 1, figures 2(c) and (f), and see figuresS1 and S2 available at stacks.iop.org/ERL/9/024005/mmedia).We include normalization by both Historical and Preindustrial

3

Environ. Res. Lett. 9 (2014) 024005 K E McCusker et al

Table 1. Shutoff SAT trends. Shutoff summer, winter, and annualaverage 5- and 20-year SAT trends (�C/decade; leftmost twocolumns), and normalized trends (rightmost two columns) in whichthe units are standard deviations (SD) of the respective trenddistribution from the Historical simulations, ‘Hist’, and from thePreindustrial simulation, ‘Prei’ in parentheses.

5yr 20yr 5yr Hist (Prei) SD 20yr Hist (Prei) SD

Summer land 3.3 1.3 1.3 (1.4) 4.5 (5.1)ocean 2.1 0.9 1.4 (1.5) 5.5 (5.7)global 2.4 1.0 1.4 (1.5) 4.9 (5.5)

Winter land 4.2 1.4 1.0 (1.0) 3.2 (3.5)ocean 2.9 1.1 1.4 (1.4) 4.6 (5.1)global 3.2 1.2 1.3 (1.3) 4.2 (4.6)

ANN land 3.6 1.4 1.8 (1.9) 5.6 (6.8)ocean 2.7 1.1 2.0 (2.1) 6.4 (7.3)global 3.0 1.2 1.9 (2.1) 6.2 (7.2)

standard deviation for reference, but focus our analysis mostlyon the Historical normalization.

Not only are the trends following SRM cessation largecompared to 20th century or preindustrial climate variability,their spatial extensiveness is also unmatched historically.Historical and Preindustrial land SAT trends exhibit a Gaussianshape centered approximately about 0 �C/decade, indicatingthat at any given time, about 50% of regional trends wouldbe warming and 50% cooling, with the trends making up thetails occurring rarely (figure S4(a) available at stacks.iop.org/ERL/9/024005/mmedia). Five-year trends especially can bevery large (>8 �C/decade; figure S4(a) available at stacks.iop.org/ERL/9/024005/mmedia); approaching the magnitudeof regional trends following SRM cessation (figure 2(a)).Note that the distribution of Historical 20-year trends isshifted slightly toward more warming trends compared to thePreindustrial control, due to the influence of increasing GHGconcentrations.

To further quantify the spatial extensiveness of land SATtrends following SRM cessation, we calculate the probabilitydensity distribution across all (area weighted) land grid cells ineach ensemble member, for the summer SAT trends followingcessation normalized by both the Historical and Preindustrialstandard deviations, and their cumulative density distributions(figure S5(a) available at stacks.iop.org/ERL/9/024005/mmedia). Cessation of SRM greatly increases the probability of‘extreme’ (i.e., trends greater than 2 standard deviations ofpreindustrial or historical trends; figure S3 available at stacks.iop.org/ERL/9/024005/mmedia) warming trends over land(figure S5(a) available at stacks.iop.org/ERL/9/024005/mmedia). After SRM termination, the probability of extreme, relativeto either 20th century or preindustrial variability, 5-year and20-year summer trends occurring somewhere over land isabout 15% and 75%, respectively. Twenty-year summer landtrends exceeding 5 standard deviations of historical variabilityhave a probability of nearly 20% (figure S5(a) available at stacks.iop.org/ERL/9/024005/mmedia), and the largest trendstend to occur in the already stressed, less resilient regions inthe tropics (figures 2(e) and (f)).

Also important for food production and the ability ofecosystems to withstand rapid heating is the amount ofprecipitation falling locally—more precipitation could leadto a greater ability of crops and ecosystems to avoid heatstress due to warming. Global average precipitation, unlikeglobal average SAT, is reduced well below Preindustrial valuesduring the period of sulfate geoengineering prior to termination(figure 1(b)) as predicted by Bala et al (2008). Upon SRMcessation, global average precipitation rapidly increases asexpected (figure 1(b)), and is consistent with the recent resultsof Jones et al (2013).

One might expect there to be greater precipitation onland following SRM cessation due to the land warmingfaster than the ocean (table 1), which would induce in-creased monsoon circulations. We find, however, that average5-year summer precipitation trends on land are negative(�0.17 mm day�1/decade compared with +0.43 mm day�1/decade over ocean), indicating general drying that would ex-acerbate the effects of rapid warming. As with the 5-year landSAT trends, the pattern of 5-year precipitation trends showsgreat spatial variation (figure 3(a)). In particular, the SoutheastAsian monsoon region shows rapidly increasing precipitation,greater than 8 mm day�1/decade, and other regions such asNortheast South America, Saudi Arabia, and India experiencelarge drying up to �8 mm day�1/decade. Nevertheless, thisspatial pattern does not persist over 20-year trends, whereinstead the Indian monsoon shows strengthening and Brazilreceives increasing precipitation (figure 3(d)). These trends,especially 5-year trends, are within 2 standard deviations ofHistorical and Preindustrial variability nearly everywhere overland (figures 3(b), (e), and figure S5(b) available at stacks.iop.org/ERL/9/024005/mmedia). Global precipitation trends arelargest over tropical oceans (figure S6 available at stacks.iop.org/ERL/9/024005/mmedia), but, unlike SAT, the variabilityin 5- and 20-year trends is also largest over oceans (figure S7available at stacks.iop.org/ERL/9/024005/mmedia).

Perhaps more important for ecosystems and agriculture ishow SAT and precipitation change together in a given location.In fact, 49% of the land grid boxes that experience warmingtrends over a five-year period also experience a drying trend,and 15% of those grid boxes experience extreme warmingcompared to Historical variability. Forty-six percent of theland grid boxes experience both 20-year warming trends and20-year drying trends, with almost all locations that experiencedrying trends also experiencing extreme warming. Coherentspatial relationships between the temperature and precipitationtrends are difficult to pick out, but in general, SAT trends aremore anomalous than precipitation trends in a given region,especially over 20 years (with the exception of high latitudeland; figure S8 available at stacks.iop.org/ERL/9/024005/mmedia), and few regions have SAT and precipitation trendsthat are both greater than 2 standard deviations. Thus we findthat in some regions, the strong positive 20-year warmingtrends are accompanied by negative trends in precipitation thatcould compound problems for food production or ecosystemhealth. However, precipitation patterns should be interpretedwith caution. Robust patterns of precipitation change remainelusive in multi-model comparisons of future climate change

4

Environ. Res. Lett. 9 (2014) 024005 K E McCusker et al

Figure 3. (a) The ensemble average summer 5-year precipitation trends (mm day�1/decade) for the period following SRM termination inthe Shutoff scenario. (b) The 5-year trends shown in (a), but normalized by the standard deviation of Historical 5-year trends at each gridpoint. (c) The 5-year trends shown in (a), but normalized by the standard deviation of Preindustrial 5-year trends at each grid point. ((d)–(f))are as ((a)–(c)), but for 20-year trends. The white stripe at the equator indicates the discontinuity in season from northern hemisphere tosouthern hemisphere (JJA and DJF, respectively). Trends over the ocean are shown in figure S6 (available at stacks.iop.org/ERL/9/024005/mmedia).

(Meehl et al 2007), solar geoengineering (Kravitz et al 2013),and even following termination of SRM (Jones et al 2013).Thus, it is not possible to determine with confidence theprecise regions where positive temperature trends will becountered by positive precipitation trends, because there arelarge precipitation trends associated with natural variability,especially in the tropics, as well as model biases.

The above results show that temperature trends followingcessation of SRM could far exceed the familiar bounds of20th century temperature trends, particularly over land withinthe low latitudes. Twenty-year temperature trends in particularwere found to be highly anomalous, with rates of warmingexceeding several degrees per decade, over a very broad regioncovering the low to mid latitude land masses (figures 2(e)and (f)). These results follow from only a few key aspectsof climate: the rapid increase in radiative forcing that wouldoccur upon SRM cessation; the rapid adjustment timescalesof the ‘fast’ components of the climate system, such as land(Held et al 2010); and the relatively small climate variabilityof the past century, particularly within the tropics. While thesegeneral results are robust across a range of SRM cessationscenarios, the details depend, to some extent, on the climatesensitivity of the GCM we have used, the background GHGemissions scenario we have employed, and our assumptionsabout the timing of the SRM termination. We thus shift ourfocus to an evaluation of the degree to which the global andannual mean climate response to SRM cessation is sensitiveto these assumptions.

3. Sensitivity to termination year, emissions, and

climate sensitivity

Current best estimates of climate sensitivity constrain its valueto likely be between 2 and 4.5 �C and very likely exceed

1.5 �C (Meehl et al 2007). We thus consider here a climatesensitivity range of 1.5–10 �C (see supplementary materialsfor more details available at stacks.iop.org/ERL/9/024005/mmedia), and further explore a variety of plausible backgroundGHG emissions scenarios and SRM termination years. Weemploy a simplified, one-dimensional, climate model that hasan upwelling-diffusion ocean and energy balance atmospherewith adjustable climate sensitivity (UD-EBM; from Baker andRoe (2009), similar in form to that in Hoffert et al (1980)).When tuned to capture the annual and global mean responseof CCSM4—including its equilibrium climate sensitivity of3.2 �C and transient climate response of 1.7 �C (Bitz et al

2012)—the UD-EBM successfully reproduces CCSM4’s SATtrends following SRM cessation (black symbols in figure 5).

The background radiative forcing (RF) within the UD-EBM is prescribed to follow RCP8.5 (Riahi et al 2007)and RCP2.6 (van Vuuren et al 2007) emissions scenarios,representing ‘business-as-usual’ emissions and strong GHGmitigation, respectively. SRM is simulated by maintaining RFat year 2000 levels until the time of abrupt SRM termination,at which point the RF is set to that of the background GHGemissions scenario until year 2100. Each of these simulationsis performed for the range of climate sensitivities describedabove and for a variety of shutoff years.

We first consider the case in which climate sensitivityis set to that of CCSM4 and SRM is employed to mask thebusiness-as-usual emissions scenario (RCP8.5, as was used inthe CCSM4 experiments in figures 1 and 2). When SRM isterminated following a 20-year implementation period (year2020 in figure 4; blue curve), RF abruptly increases by about1 W m2, producing a small spike in the rate of temperaturechange that quickly decays to the rate of the backgroundRCP8.5 scenario. The 20-year temperature trends followingSRM cessation are 0.2–0.6 �C/decade for the range of climate

5

Environ. Res. Lett. 9 (2014) 024005 K E McCusker et al

Figure 4. (a) Evolution of the net radiative forcing (RF) due to GHGs and SRM, (b) temperature response (4T ), and (c) rate of temperaturechange (d4T/dt) with climate sensitivity set to 3.2 �C, that of CCSM4. The business-as-usual emissions scenario (RCP8.5) andlow-emissions scenario (RCP2.6) are shown in thick, lightly shaded curves (light pink and gray, respectively). SRM termination following20 years and 80 years of implementation are shown in thinner, dark curves (blue and green for RCP8.5 and RCP2.6 background emissions,respectively). Figure S9 (available at stacks.iop.org/ERL/9/024005/mmedia) displays these results for the ‘very likely’ range of climatesensitivities defined in the Intergovernmental Panel on Climate Change (IPCC) Assessment Report 4 (AR4).

sensitivities (figure 5), comparable to those trends that occurunder the RCP8.5 scenario without any SRM.

In contrast, when SRM is implemented for a period of80 years before cessation, there is an abrupt RF increaseof over 5 W m2 (year 2080 in figure 4(a); blue curve) dueto the loss of the large SRM RF that was required to maskthe ongoing accumulation of GHGs in the atmosphere. Thisspike in RF produces a rapid and substantial increase in globalaveraged temperature: almost 9 �C/decade in the first fewyears for CCSM4’s climate sensitivity (figure 4(c)), and up to10 �C/decade for high climate sensitivities (figure S9 availableat stacks.iop.org/ERL/9/024005/mmedia). Sensitivity of ini-tial rates of change are consistent with previous evaluations ofSRM termination with multiple climate sensitivities and/or ter-mination year (Matthews et al 2007, Ross and Matthews 2009).Twenty-year trends over the range of climate sensitivities are0.6–2 �C/decade (figure 5). Thus, under business-as-usualfuture GHG emissions, the stabilization of climate with SRMfor a period of longer than about two decades would createthe potential for sustained high rates of warming upon SRMcessation, even if climate sensitivity were near the lower endof its estimated range (figure 5).

We next consider the case where SRM is employedalong with concurrent aggressive GHG mitigation measures,as represented by the low-emissions RCP2.6 scenario whereinanthropogenic RF is about 2.6 W m2 above preindustrial in2100 (Moss et al 2010). Due to the limited accumulation ofGHGs in the atmosphere, the SRM RF required to stabilizeclimate is relatively small (compared to the RCP8.5 case),and thus SRM termination results in an abrupt RF increase ofless than about 2 W m2 regardless of its timing (figure 4(a);

green curves). Following SRM cessation, there are high ratesof temperature change in the first few years (figure 4(c); greencurves), but 20-year temperature trends remain below about0.4 �C/decade—comparable to those trends that occur underthe RCP2.6 scenario without any SRM—over the full range ofclimate sensitivity and timing of SRM termination (figure 5).

Within each of the above scenarios, the initial rate oftemperature change following SRM cessation depends onclimate sensitivity only nominally (see supplementary noteand figure S10 available at stacks.iop.org/ERL/9/024005/mmedia). Climate sensitivity does become an important factor insetting longer-term temperature trends, particularly under alarge RF increase (compare 20-year trends in figure 5 with 5-year trends in figure S10 available at stacks.iop.org/ERL/9/024005/mmedia). However, figure 5 (and figure S10 available at stacks.iop.org/ERL/9/024005/mmedia) shows that the principalcontrol on the rate of temperature change following SRMcessation is the size of the abrupt RF increase, which, in turn, isdetermined jointly by the background GHG emissions scenarioand the duration of time that SRM has been deployed.

Critically then, even for the lowest plausible valuesof climate sensitivity, decadal temperature trends would beextremely large (double that of the largest 20-year historicaltrend in CCSM4; horizontal black line in figure 5) in the eventof a late 21st century SRM termination under high (RCP8.5)future emissions (figure 5; blue asterisks); conversely, evenfor the highest plausible values of climate sensitivity, decadaltemperature trends would remain relatively small in the eventof SRM cessation at any point in the 21st century, under low(RCP2.6) future emissions (figure 5; green asterisks). Thus, theonly way to avoid creating the risk of substantial temperature

6

Environ. Res. Lett. 9 (2014) 024005 K E McCusker et al

Figure 5. Twenty-year temperature trend following SRMtermination — after 20, 50, and 80 years (boxed numbers) ofimplementation — as a function of climate sensitivity for theRCP8.5 (blue asterisks) and RCP2.6 (green asterisks) backgroundemissions scenarios and the maximum RCP8.5 (blue squares) andRCP2.6 (green squares) 20-year trends. Background shadingindicates the Intergovernmental Panel on Climate Change (IPCC)Assessment Report 4 (AR4) likelihood ranges for climate sensitivity(‘likely’ has a >66% probability, and ‘very likely’ has a >90%probability; Hegerl et al 2007). The vertical black line is CCSM4’sclimate sensitivity (3.2 �C), the horizontal black line is themaximum global mean, annual mean 20-year trend sampled fromthe CCSM4 Historical simulations, the black triangle shows theCCSM4 20-year global mean, annual mean trend following SRMcessation (1.16 �C/decade), and the black circle shows the 20-yearUD-EBM trend for SRM termination after 65 years of balanced RF(1.0 �C/decade), when the RF jump is roughly equivalent to thatestimated in CCSM4 (about 4–5W m�2).

trends through SRM is concurrent strong reductions of GHGemissions.

While we have considered here only SRM cessationscenarios within the 21st century, these findings have long-term implications as well. If SRM was used to stabilize globalclimate under high future GHG emissions, it would need to bemaintained on timescales determined by the turnover time ofGHGs in the atmosphere, which in the case of carbon dioxideis multiple millennia (Archer 2005). Indeed, the stabilizationof global temperature with SRM would also preclude furtherobservations of the climate response to the ongoing GHGemissions (Matthews et al 2007), on which many estimates ofglobal climate sensitivity are based. Thus, the large-scale useof SRM to mask business-as-usual GHG emissions could leadto a scenario wherein SRM must be maintained for millennia,else risking a large and uncertain level of rapid global warmingupon any unanticipated cessation.

4. Discussion and conclusions

Previous studies have identified the potential for rapid anddangerous global warming following the cessation of SRM(Wigley 2006, Matthews et al 2007, Ross and Matthews2009, Brovkin et al 2009, Robock et al 2008, McCusker

et al 2012, Irvine et al 2012, Jones et al 2013). The re-sults presented here reinforce and extend this assessment by(i) quantifying the regional and seasonal climate responseto SRM cessation, of critical importance for the impacts onecological and human systems, and (ii) demonstrating thatover a wide range of plausible 21st century scenarios, ratesof warming are primarily controlled by the accumulated GHGemissions that are abruptly unmasked upon SRM cessation.Given unabated emissions, the spatial and temporal extent ofSAT trends caused by a cessation of SRM would be wellbeyond the bounds experienced in the last century, and wouldfar exceed those considered safe for many ecological systems(van Vliet and Leemans 2006, Lenton 2011). The spatialstructure of the trends precludes the possibility that therewill be isolated regions that may experience brief periodsof ‘relief’ from high rates of warming. Moreover, greaterthan 40% of the land area that experiences warming trendsdue to SRM cessation also experiences drying trends. Thus,food production could be severely reduced in many regionsconcurrently under a scenario of high GHG emissions andSRM termination. Furthermore, the adaptive options of manyspecies reach their limits under standard projected climatechanges, let alone the widespread and rapid changes that couldoccur due to SRM cessation. Finally, one potential positive toSRM cessation for photosynthesizing organisms in particular,is increased net primary productivity on land (figures S9and S10 available at stacks.iop.org/ERL/9/024005/mmedia),however disagreement among global climate models on thesign of the response to cessation (Jones et al 2013) leavesthis an open question (see supplementary materials for furtherdetails available at stacks.iop.org/ERL/9/024005/mmedia).

Alternative climate change mitigation measures couldarguably become necessary should climate change progressat a rate or to a degree deemed dangerous to ecologicalor human systems. Such a scenario could arise if GHGemissions continue unabated, or if climate sensitivity is higherthan anticipated. While it has been argued that SRM wouldbe particularly effective in curbing future climate changeunder high emissions or high climate sensitivity (Ricke et al

2012), our results show that the warming following SRMcessation becomes most severe under these same conditions.We are thus left with the disconcerting situation in whichSRM is most useful precisely when its associated risks arethe greatest. Furthermore, SRM via stratospheric aerosolsmay introduce a host of additional problems, among themchanges in atmospheric and oceanic circulations that act todestabilize the West Antarctic ice sheet (McCusker et al 2012)and stratospheric ozone depletion (Tilmes et al 2008). It hasbeen suggested that SRM be combined with GHG emissionsmitigation with the aim of simultaneously limiting globalwarming and ocean acidification (Wigley 2006). Our resultsemphasize that should SRM ever be implemented, aggressiveemissions mitigation must occur simultaneously due to theclimatic risks involved with its abrupt cessation.

Acknowledgments

The authors wish to thank two anonymous reviewers forhelping to substantially improve the manuscript. This research

7

Environ. Res. Lett. 9 (2014) 024005 K E McCusker et al

was funded by the Tamaki Foundation and supported in part bythe National Science Foundation through TeraGrid resourcesprovided by the Texas Advanced Computing Center underGrant TG-ATM090059. KCA received support from a JamesS McDonnell Foundation Postdoctoral Fellowship.

References

Ammann C M, Washington W M, Meehl G A, Buja L and Teng H2010 Climate engineering through artificial enhancement ofnatural forcings: magnitudes and implied consequencesJ. Geophys. Res. 115 D22109

Archer D 2005 Fate of fossil fuel CO2 in geologic time J. Geophys.

Res. 110 C09S05Baker M B and Roe G H 2009 The shape of things to come: why is

climate change so predictable? J. Clim. 22 4574–89Bala G, Duffy P B and Taylor K E 2008 Impact of geoengineering

schemes on the global hydrological cycle Proc. Natl Acad. Sci.

105 7664–9Battisti D S and Naylor R L 2009 Historical warnings of future food

insecurity with unprecedented seasonal heat Science 323 240–4Bell G and Collins S 2008 Adaptation, extinction and global change

Evol. Appl. 1 3–16Bitz C M et al 2012 Climate sensitivity of the community climate

system model, version 4 J. Clim. 25 3053–70Brovkin V, Petoukhov V, Claussen M, Bauer E, Archer D and

Jaeger C 2009 Geoengineering climate by stratospheric sulfurinjections: earth system vulnerability to technological failureClim. Change 92 243–59

FAO, WFP and IFAD 2012 The State of Food Insecurity in theWorld 2012. Economic growth is necessary but not sufficient toaccelerate reduction of hunger and malnutrition. Rome, FAOwww.fao.org/docrep/016/i3027e/i3027e00.pdf

Gent P R et al 2011 The community climate system model version 4J. Clim. 24 4973–91

Hansen J et al 2006 Global temperature change Proc. Natl Acad.

Sci. USA 103 14288–93Hansen J, Sato M, Kharecha P and von Schuckmann K 2011 Earth’s

energy imbalance and implications Atmos. Chem. Phys. Discuss.

11 27031–105Hegerl G C et al 2007 Climate Change 2007: The Physical Science

Basis. Contribution of Working Group I to the Fourth Assessment

Report of the Intergovernmental Panel on Climate Change ed SSolomon et al (Cambridge: Cambridge University Press)pp 663–745

Held I M et al 2010 Probing the fast and slow components of globalwarming by returning abruptly to preindustrial forcing J. Clim.

23 2418–27Hoffert M I, Callegari A J and Hsieh C-T 1980 The role of deep sea

heat storage in the secular response to climatic forcingJ. Geophys. Res. 85 6667–79

Irvine P J, Sriver R L and Keller K 2012 Tension between reducingsea-level rise and global warming through solar-radiationmanagement Nature Clim. Change 2 97–100

Jones A et al 2013 The impact of abrupt suspension of solarradiation management (termination effect) in experiment G2 ofthe Geoengineering Model Intercomparison Project (GeoMIP)J. Geophys. Res. 118 9743–52

Keith D W, Parson E and Morgan M G 2010 Research on global sunblock needed now Nature 463 426–7

Kravitz B et al 2013 Climate model response from thegeoengineering model inter-comparison project (GeoMIP)J. Geophys. Res.: Atmos. 118 8320–32

Lenton T M 2011 Beyond 2 �C: redefining dangerous climatechange for physical systems Wiley Interdis. Rev.: Clim. Change

2 45146Lobell D B, Schlenker W and Costa-Roberts J 2011 Climate trends

and global crop production since 1980 Science 233 616–20Lyman J M et al 2010 Robust warming of the global upper ocean

Nature 465 334–7Matthews H D and Caldeira K 2007 Transient climate-carbon

simulations of planetary geoengineering Proc. Natl Acad. Sci.

USA 104 9949–54McCusker K E, Battisti D S and Bitz C M 2012 The climate

response to stratospheric sulfate injections and implications foraddressing climate emergencies J. Clim. 25 3096–116

Meehl G A et al 2007 Climate Change 2007: The Physical Science

Basis. Contribution of Working Group I to the Fourth Assessment

Report of the Intergovernmental Panel on Climate Change ed SSolomon et al (Cambridge: Cambridge University Press)pp 747–846

Moss R H et al 2010 The next generation of scenarios for climatechange research and assessment Nature 463 747–56

Rasch P J, Crutzen P J and Coleman D B 2008 Exploring thegeoengineering of climate using stratospheric sulfate aerosols:The role of particle size Geophys. Res. Lett. 35 L02809

Rasch P J et al 2008 An overview of geoengineering of climateusing stratospheric sulphate aerosols Philos. Trans. Roy. Soc. A366 4007–37

Riahi K, Gruebler A and Nakicenovic N 2007 Scenarios oflong-term socio-economic and environmental development underclimate stabilization Technol. Forecast. Soc. Change 74 887–935

Ricke K L, Rowlands D J, Ingram W J, Keith D W andMorgan M G 2012 Effectiveness of stratospheric solar-radiationmanagement as a function of climate sensitivity Nature Clim.

Change 2 92–6Robock A 2000 Volcanic eruptions and climate Rev. Geophys.

38 191–219Robock A, Oman L and Stenchikov G L 2008 Regional climate

responses to geoengineering with tropical and Arctic SO2injections J. Geophys. Res. 113 D16101

Ross A and Matthews H D 2009 Climate engineering and the risk ofrapid climate change Environ. Res. Lett. 4 045103

Schloss C A, Nunez T A and Lawler J J 2012 Dispersal will limitability of mammals to track climate change in the WesternHemisphere Proc. Natl Acad. Sci. USA 109 8606–11

Tilmes S, Muller R and Salawitch R 2008 The sensitivity of polarozone depletion to proposed geoengineering schemes Science

320 1201–4van Vliet A and Leemans R 2006 Rapid species responses to

changes in climate require stringent climate protection targetsAvoiding Dangerous Climate Change ed H J Schellnhuber,W Cramer, N Nakicenovic, T Wigley and G Yohe (Cambridge:Cambridge University Press) pp 135–41

van Vuuren D P et al 2007 Stabilizing greenhouse gasconcentrations at low levels: an assessment of reductionstrategies and costs Clim. Change 81 119–59

Vaughan N E and Lenton T M 2011 A review of climategeoengineering proposals Clim. Change 109 745–90

Wigley T M L 2006 A combined mitigation/geoengineeringapproach to climate stabilization Science 314 452–4

8